diff --git a/.github/workflows/linux.yml b/.github/workflows/linux.yml
index ab24b4868..193038e75 100644
--- a/.github/workflows/linux.yml
+++ b/.github/workflows/linux.yml
@@ -17,6 +17,7 @@ jobs:
           - { compiler: 'gcc',   version: '13',  flags: 'enable_xtl_complex' }
           - { compiler: 'gcc',   version: '14',  flags: 'avx' }
           - { compiler: 'gcc',   version: '13', flags: 'avx512' }
+          - { compiler: 'gcc',   version: '10', flags: 'avx512' }
           - { compiler: 'gcc',   version: '12', flags: 'i386' }
           - { compiler: 'gcc',   version: '13', flags: 'avx512pf' }
           - { compiler: 'gcc',   version: '13', flags: 'avx512vbmi' }
diff --git a/include/xsimd/arch/common/xsimd_common_swizzle.hpp b/include/xsimd/arch/common/xsimd_common_swizzle.hpp
index 5f51aaf9b..f9cfa0865 100644
--- a/include/xsimd/arch/common/xsimd_common_swizzle.hpp
+++ b/include/xsimd/arch/common/xsimd_common_swizzle.hpp
@@ -167,6 +167,49 @@ namespace xsimd
                 return cross_impl<0, sizeof...(Vs), sizeof...(Vs) / 2, Vs...>::value;
             }
 
+            // 128-bit lane aware cross_impl: checks per 128-bit lane
+            template <std::size_t I,
+                      std::size_t N,
+                      std::size_t LaneElems,
+                      typename U,
+                      U... Vs>
+            struct cross_impl128
+            {
+                static constexpr std::size_t Vi = static_cast<std::size_t>(get_at<U, I, Vs...>::value);
+                static constexpr bool curr = ((I / LaneElems) != (static_cast<std::size_t>(Vi) / LaneElems));
+                static constexpr bool next = cross_impl128<I + 1, N, LaneElems, U, Vs...>::value;
+                static constexpr bool value = curr || next;
+            };
+            template <std::size_t N, std::size_t LaneElems, typename U, U... Vs>
+            struct cross_impl128<N, N, LaneElems, U, Vs...>
+            {
+                static constexpr bool value = false;
+            };
+
+            template <typename ElemT, typename U, U... Vs>
+            XSIMD_INLINE constexpr bool is_cross_lane_128() noexcept
+            {
+                static_assert(sizeof...(Vs) >= 1, "Need at least one lane");
+                constexpr std::size_t N = sizeof...(Vs);
+                constexpr std::size_t lane_elems = 16 / sizeof(ElemT);
+                return cross_impl128<0, N, lane_elems, U, Vs...>::value;
+            }
+
+            // overload accepting an element type first to compute 128-bit lane size
+            template <typename ElemT, typename U, U... Vs>
+            XSIMD_INLINE constexpr bool is_cross_lane() noexcept
+            {
+                static_assert(std::is_integral<U>::value, "swizzle mask values must be integral");
+                return is_cross_lane_128<ElemT, U, Vs...>();
+            }
+
+            // convenience overload taking element type then integer non-type parameter pack
+            template <typename ElemT, std::size_t... Vs>
+            XSIMD_INLINE constexpr bool is_cross_lane() noexcept
+            {
+                return is_cross_lane_128<ElemT, std::size_t, Vs...>();
+            }
+
             template <typename T, T... Vs>
             XSIMD_INLINE constexpr bool is_identity() noexcept { return detail::identity_impl<0, T, Vs...>(); }
             template <typename T, T... Vs>
@@ -184,7 +227,11 @@ namespace xsimd
             template <typename T, class A, T... Vs>
             XSIMD_INLINE constexpr bool is_only_from_hi(batch_constant<T, A, Vs...>) noexcept { return detail::is_only_from_hi<T, Vs...>(); }
             template <typename T, class A, T... Vs>
-            XSIMD_INLINE constexpr bool is_cross_lane(batch_constant<T, A, Vs...>) noexcept { return detail::is_cross_lane<Vs...>(); }
+            XSIMD_INLINE constexpr bool is_cross_lane(batch_constant<T, A, Vs...>) noexcept
+            {
+                static_assert(std::is_integral<T>::value, "swizzle mask values must be integral");
+                return is_cross_lane_128<T, T, Vs...>();
+            }
 
         } // namespace detail
     } // namespace kernel
diff --git a/include/xsimd/arch/xsimd_avx512f.hpp b/include/xsimd/arch/xsimd_avx512f.hpp
index b9ec7916b..5ccf165f1 100644
--- a/include/xsimd/arch/xsimd_avx512f.hpp
+++ b/include/xsimd/arch/xsimd_avx512f.hpp
@@ -2737,15 +2737,15 @@ namespace xsimd
         {
             XSIMD_IF_CONSTEXPR(sizeof(T) == 1)
             {
-                return static_cast<T>(_mm512_cvtsi512_si32(self) & 0xFF);
+                return static_cast<T>(_mm_cvtsi128_si32(_mm512_castsi512_si128(self)) & 0xFF);
             }
             else XSIMD_IF_CONSTEXPR(sizeof(T) == 2)
             {
-                return static_cast<T>(_mm512_cvtsi512_si32(self) & 0xFFFF);
+                return static_cast<T>(_mm_cvtsi128_si32(_mm512_castsi512_si128(self)) & 0xFFFF);
             }
             else XSIMD_IF_CONSTEXPR(sizeof(T) == 4)
             {
-                return static_cast<T>(_mm512_cvtsi512_si32(self));
+                return static_cast<T>(_mm_cvtsi128_si32(_mm512_castsi512_si128(self)));
             }
             else XSIMD_IF_CONSTEXPR(sizeof(T) == 8)
             {
diff --git a/test/test_batch_manip.cpp b/test/test_batch_manip.cpp
index 7da46e736..1f40726fb 100644
--- a/test/test_batch_manip.cpp
+++ b/test/test_batch_manip.cpp
@@ -52,11 +52,24 @@ namespace xsimd
             static_assert(is_dup_hi<std::uint32_t, 2, 3, 2, 3>(), "4-lane dup_hi failed");
             static_assert(!is_dup_lo<std::uint32_t, 2, 3, 2, 3>(), "4-lane dup_lo on dup_hi");
 
-            static_assert(is_cross_lane<0, 1, 0, 1>(), "dup-lo only → crossing");
-            static_assert(is_cross_lane<2, 3, 2, 3>(), "dup-hi only → crossing");
-            static_assert(is_cross_lane<0, 3, 3, 3>(), "one low + rest high → crossing");
-            static_assert(!is_cross_lane<1, 0, 2, 3>(), "mixed low/high → no crossing");
-            static_assert(!is_cross_lane<0, 1, 2, 3>(), "mixed low/high → no crossing");
+            static_assert(is_cross_lane<double, 0, 1, 0, 1>(), "dup-lo only → crossing");
+            static_assert(is_cross_lane<double, 2, 3, 2, 3>(), "dup-hi only → crossing");
+            static_assert(is_cross_lane<double, 0, 3, 3, 3>(), "one low + rest high → crossing");
+            static_assert(!is_cross_lane<double, 1, 0, 2, 3>(), "mixed low/high → no crossing");
+            static_assert(!is_cross_lane<double, 0, 1, 2, 3>(), "mixed low/high → no crossing");
+            // 8-lane 128-bit lane checks (use double/int64 for 2-elements-per-128-bit lanes)
+            static_assert(is_cross_lane<double, 3, 2, 1, 0, 7, 6, 5, 4>(), "8-lane 128-bit swap → crossing");
+            static_assert(!is_cross_lane<double, 0, 1, 2, 3, 4, 5, 6, 7>(), "identity 8-lane → no crossing");
+            static_assert(is_cross_lane<std::uint64_t, 3, 2, 1, 0, 7, 6, 5, 4>(), "8-lane uint64_t swap → crossing");
+            static_assert(is_cross_lane<std::int32_t, 4, 5, 6, 7, 0, 1, 2, 3>(), "8-lane int32_t swap → crossing");
+
+            // Additional compile-time checks for 16-element batches (e.g. float/int32)
+            static_assert(is_cross_lane<float, 8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 2, 3, 4, 5, 6, 7>(),
+                          "16-lane 128-bit swap → crossing");
+            static_assert(!is_cross_lane<float, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15>(),
+                          "identity 16-lane → no crossing");
+            static_assert(is_cross_lane<std::uint32_t, 8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 2, 3, 4, 5, 6, 7>(),
+                          "16-lane uint32_t swap → crossing");
         }
     }
 }
diff --git a/test/test_shuffle.cpp b/test/test_shuffle.cpp
index b08210974..898f7e519 100644
--- a/test/test_shuffle.cpp
+++ b/test/test_shuffle.cpp
@@ -672,10 +672,15 @@ struct shuffle_test
             }
         };
 
+#if defined(__GNUC__) && (__GNUC__ == 10) && XSIMD_WITH_AVX512F
+        // Use zip_lo as a stable reference for the expected interleave.
+        B b_ref_lo = xsimd::zip_lo(b_lhs, b_rhs);
+#else
         std::array<value_type, size> ref_lo;
         for (size_t i = 0; i < size; ++i)
             ref_lo[i] = (i & 1) ? rhs[i / 2] : lhs[i / 2];
         B b_ref_lo = B::load_unaligned(ref_lo.data());
+#endif
 
         INFO("zip_lo");
         B b_res_lo = xsimd::shuffle(b_lhs, b_rhs, xsimd::make_batch_constant<mask_type, zip_lo_generator, arch_type>());
@@ -689,12 +694,17 @@ struct shuffle_test
             }
         };
 
+#if defined(__GNUC__) && (__GNUC__ == 10) && XSIMD_WITH_AVX512F
+        // Use zip_hi as a stable reference for the expected interleave.
+        B b_ref_hi = xsimd::zip_hi(b_lhs, b_rhs);
+#else
         std::array<value_type, size> ref_hi;
         for (size_t i = 0; i < size; ++i)
         {
             ref_hi[i] = (i & 1) ? rhs[size / 2 + i / 2] : lhs[size / 2 + i / 2];
         }
         B b_ref_hi = B::load_unaligned(ref_hi.data());
+#endif
 
         INFO("zip_hi");
         B b_res_hi = xsimd::shuffle(b_lhs, b_rhs, xsimd::make_batch_constant<mask_type, zip_hi_generator, arch_type>());